The outputs from Elude using Linux Ubuntu (elude-v3.03) and Windows x64 (elude-v3.03) are significantly different. The output from Linux seems in line with expected iRT values (as the model was trained to output iRT values) and has a good linear correlation with measured RT values, whereas the Windows output does not align with any expected value (outputs on the order of >1e8). I've seen nearly identical results using different models (including those supplied with Elude), different versions of elude (v3.03 and v3.04), and using the -n argument (with and without).

Is there a reason for the difference in outputs, and if so, a way to readily convert the Windows output to what I'm seeing for the linux output? Thanks.

Example Linux Output

# File generated by Elude. Predicted retention times are given below.
# Jul 23 2019 , 14:10:33
Peptide Predicted_RT
AAFGGSGGRGSSSGGGYSSGSSSYGSGGR 31.4642
ADLEMQIESLKEELAYLK 103.11
ADLEMQIESLTEELAYLK 116.278
ADLEMQIESLTEELAYLKK 104.975
ADLTGISPSPNLYLSK 69.7456
AEDGSVIDYELIDQDAR 69.5236

Example Windows x64 Output

# File generated by Elude. Predicted retention times are given below.
# Jul 23 2019 , 07:36:53
Peptide Predicted_RT
AAFGGSGGRGSSSGGGYSSGSSSYGSGGR 2.55902e+08
ADLEMQIESLKEELAYLK 2.2453e+08
ADLEMQIESLTEELAYLK 2.38189e+08
ADLEMQIESLTEELAYLKK 2.20331e+08
ADLTGISPSPNLYLSK 2.69071e+08
AEDGSVIDYELIDQDAR 3.10754e+08

• elude: v3.03
• windows: Windows 10 Enterprise (1803.17134.1009)
• linux: Ubuntu 18.04.1 (Bionic)
• model: https://github.com/brvpuyve/MS2PIP-for-DIA/blob/master/ELUDE/ELUDE_PAN_HUMAN.model
• example command:

elude -l ELUDE_PAN_HUMAN.model -e tmp_elude_input.txt -o tmp_eludePredictions_win64-v3-03.txt -p -v 5

tmp_elude_input.txt
tmp_eludePredictions_linuxUbuntu_v3-03.txt
tmp_eludePredictions_win64_v3-03.txt